Quantopian's community platform is shutting down. Please read this post for more information and download your code.
Back to Community
Algorithms for Small Investment Amounts

I'm a college student with over 9 years of programming experience and a big interest in machine learning and trading.

I have about $4000-5000 of disposable money that I would like to play around with in the context of algorithmic trading. I am wondering what class or types of algorithms would be best for trading with small amounts like this (considering the cost of commissions) ? Of course I will be backtesting and paper trading any algorithms I decide on before actually putting my money at risk.

Also, I am curious as to why people on here are so open about the algorithms they have been developing. Some of these algorithms appear to be making unbelievable returns on the backtest, why wouldn't you guys just keep it a secret and reap the monetary reward? Are you more interested in the challenge of writing successful algorithms than in actually using them to make money?

10 responses

Hello Brandon,

Taking a quick look at the Interactive Brokers site (http://www.interactivebrokers.com/en/index.php?f=4969):

  • Balance - Individuals age 26 or younger: USD 3,000 (or non-USD equivalent)
  • Monthly Activity - Age 25 or under: USD 3.00 (or non-USD equivalent)

So, if I'm reading things correctly and you are 25 years old or younger, you can set up an IB account and pay a $36 per year maintenance fee (with $5000, this is 0.72% annually, which is actually quite high compared to the Vanguard's Prime Money Market expense ratio of 0.16%, for example).

I've gathered that it'll cost ~$1 per transaction through IB (to my knowledge, Quantopian has not revealed their anticipated commission structure yet, but my assumption is that they'll need to charge a commission on top of IB's).

I don't understand the implications, but you need to investigate the impact of odd lot orders (see http://ibkb.interactivebrokers.com/article/1062 for a discussion). I can't sort out if the direct trading costs are higher, or if there is just some form of slippage.

You might also consider the tax implications...looks like IB has IRA accounts.

One suggestion is to identify a long-only, buy-and-hold strategy with a portfolio that you re-balance periodically (weekly/monthly/quarterly/"seasonally"). However, you need to benchmark your strategy against a comparable low-cost mutual fund investment (e.g. with Vanguard), in an IRA. One problem I see is that you can't really build a portfolio with $5000. For example, if you have 100 shares of a stock at $50 per share, you've consumed all of your capital.

Reportedly, Forex trading can be much more profitable for small-time retail traders, compared to securities trading...perhaps somebody experienced can confirm.

Regarding the crazy returns posted on Quantopian, typically, they are due to non-existent or unrealistic borrowing limits and/or a bias (e.g. buy shares of Apple over the past 10 years). My sense is that experienced traders have not posted their profitable proprietary algorithms on Quantopian (although there have been numerous implementations of well-known/published strategies/algorithms).

Grant

As a back-of-the-envelope sanity check, let's say you actually did daily trading. As a test case, imagine re-balancing a portfolio of 5 securities every day. So, that's $5 / day * 5 days / week * 50 weeks / year = $1250 / year. As a percentage, that's ($1250 / $5000) * 100% = 25%. To be competitive with a low-cost index fund, you'd need ~100X more money, or ~$500,000.

Taking this further, say that as a estimate, you shouldn't put more than 10% of your net worth into active trading. So, you'd need a net worth of $5 million.

--Grant

If you try FX, Futures and Options it might be better use to leverage. You have to miraculously manage the risk through.

Hello Suminda,

My hunch is that $5K is gonna be too little for any kind of intraday, active trading strategy (regardless of instrument), right? And can one buy on margin with only $5K?

In my mind, there's probably something like a $50K to $100K capital range where it might start to make sense to think about active trading (even at a daily frequency). Is this a reasonable estimate?

Again, the $50K-$100K range implies a net worth of ~10X, or $500K to $1M. Unless you want to "bet the farm."

Grant

I think 10k in FX is OK to start with. In IB it does raise TC as there is no Mini and Micro lots. With 20x leverage it will be 200k.

Other instruments are SSF, CFD, etc. which has a lot of leverage.

Again the key here is meticulous risk management as losses can be a large % of capital.

Most of the backtests being posted here are unrealistic. They are tested with little or no commissions and trading with large account sizes while keeping most of the net worth of the account invested. Even a single major drop going forward will do some serious damage. This can and will be encountered in a forward test.

Thanks for the replies guys.

I figured intra-day trading would require great amounts of risk-taking to be able to just possibly break even with fees considered.
However, I am interested in news/sentiment style trading. If a news story breaks about a particular company, I want my algorithm to determine the sentiment of that story (good news, bad news, neutral, etc) and buy or sell quickly based on the assessment. This could be something that trades every few days or so, not multiple times in the same day.

First, is that even possible on quantopian? Is it possible to download, for example, source code from external sites to analyze the text? Secondly, would that strategy alleviate the problems of very short term trading with merely $5k?

Hello Brandon,

You might check out the fetcher feature (see the help docs.). I haven't used it yet, but it might work for your application. Also, there is an offline, free backtesting engine called zipline that you could run on your pc with your own data (e.g. from Yahoo). You should be able to incorporate external signals.

Grant

Brandon,

My hunch is that you are still going to have problems with only $5K...you might look into an account that offers limited trades, but they are free. For example, if you can stick to Vanguard ETFs, limited trading is free (see https://personal.vanguard.com/us/whatweoffer/stocksbondscds/feescommissions).

Some relevant links:

http://www.predictivesignals.com/

https://www.recordedfuture.com/

https://www.quantopian.com/posts/new-nature-paper-quantifying-wikipedia-usage-patterns-before-stock-market-moves

Grant

this recommendation is a relatively new thing...it might be worth looking in to it.
www.nadex.com